GPT-5.5 Bio Bug Bounty

OpenAI has introduced the GPT-5.5 Bio Bug Bounty, a targeted red-teaming challenge designed to uncover universal jailbreaks that could compromise GPT-5.5's bio safety guardrails. The program offers rewards up to $25,000 for severe, reproducible exploits that bypass the model's protections against generating harmful biological content. This initiative is part of OpenAI's broader safety strategy, focusing on high-risk domains like biosecurity where AI misuse could have catastrophic consequences.

Participants are tasked with identifying vulnerabilities that could enable the model to provide detailed instructions for creating biological threats or bypass safety filters consistently. The challenge emphasizes universal jailbreaks—exploits that work across multiple prompts and contexts—rather than isolated failures. By incentivizing external researchers to stress-test GPT-5.5's defenses, OpenAI aims to preemptively patch weaknesses before malicious actors can exploit them. The bounty reflects growing industry attention on aligning advanced AI systems with safety standards in sensitive areas like biosecurity.